To address the challenge of controlling unmanned aerial vehicle e (UAV) swarms under trajectory and detection constraints, this study proposes a closed-loop control framework for predictive compensation tracking based on historical perception data and motion models. A mathematical simulation platform was developed, alongside a task verification platform for formation control based on the leader-follower method and a visual simulation system, enabling comprehensive simulation of the UAV swarm’s process from target detection to direct and predictive tracking. Experimental results demonstrate that the framework can generate desired trajectories compliant with constraints, effectively achieving autonomous target perception and closed-loop predictive tracking control, while significantly extending the swarm’s continuous target tracking duration.